基于CUACE模式产品的广西空气质量预报效果对比分析  

Comparative Analysis of Air Quality Forecast Effect in Guangxi Based on CUACE Model Products

作　　者：廖国莲[1] 曾鹏 莫雨淳[1] 郑凤琴 Liao Guolian;Zeng Peng;Mo Yuchun;Zheng Fengqin(Guangxi Meteorological Observatory,Nanning 530022,China;Guangxi Meteorological Disaster Prevention Technology Center,Nanning 530022,China;Guangxi Climate Center,Nanning 530022,China)

机构地区：[1]广西气象台,南宁530022 [2]广西气象灾害防御技术中心,南宁530022 [3]广西气候中心,南宁530022

出　　处：《气象与环境科学》2022年第4期39-44,共6页Meteorological and Environmental Sciences

基　　金：广西自然科学基金项目(2014GXNSFBA118216);广西科技计划项目(桂科AB16380292);广西气象科研计划项目(桂气科2020Z04、2016M02)。

摘　　要：利用广西14个地市的空气质量监测资料和气象资料,通过线性回归和神经网络方法,对中国气象局下发的CUACE模式空气质量预报指导产品进行订正,并对比两种方法的预报效果。结果表明,神经网络模型对南宁各月的预报分数均在66分以上,对CUACE模式24 h预报产品的订正效果最明显,可作为南宁市空气质量预报业务的重要参考;线性回归模型对南宁各月的预报分数为55~71,分数差别较大,评分较低的月份主要出现在夏秋季节。在颗粒物(PM_(10)和PM_(2.5))浓度的预报业务中,神经网络模型预报在广西大部分城市预报误差较小,具有较高参考价值;河池和梧州可利用线性回归对O_(3)浓度进行预报,广西其他地市则可采用神经网络模型预报O3浓度。虽然CUACE模式预报仅在个别城市的个别预报时效有一定优势,但CUACE模式预报效果最稳定,其评分基本不随预报时效的延长而明显降低,在时效较长的趋势预报上,可以重点关注CUACE模式预报产品。Based on the air quality monitoring data and meteorological data of 14 cities in Guangxi Zhuang Autonomous Region,the CUACE model air quality forecast guidance products issued by China Meteorological Administration are revised by linear regression and neural network method,and the forecast effects of the two methods are compared.The results show that the monthly forecast scores of neural network model for Nanning are all above 66 points,and its correction effect to the 24 h forecast product of CUACE model is the most obvious,which can be used as an important reference for the air quality forecasting operation in Nanning.The scores of linear regression method for each month of Nanning are between 55 and 71,and there is a big difference in the scores.The low-score months are mainly in summer and autumn.In the forecasting operation of particulate matter(PM_(10) and PM_(2.5)),the neural network forecasts in most cities of Guangxi have smaller errors,so they take a high reference value.In Hechi and Wuzhou,linear regression can be used to predict O_(3) concentration,while neural network can be adopted in other cities of Guangxi to predict O_(3) concentration.Although the CUACE model has certain advantages only in some prediction lead time in individual cities,it is most stable in its forecast effect and its score does not significantly decrease with the extension of prediction lead time.Therefore,for long-term trend prediction,the CUACE model prediction products can be paid special attention to.

关 键 词：CUACE模式 空气质量 预报效果 

分 类 号：X513[环境科学与工程—环境工程] P456.8[天文地球—大气科学及气象学]